The US Centers for Disease Control initiated physician-based chronic fatigue syndrome (CFS) surveillance systems in four cities in September 1989 to determine the prevalence, incidence, course and impact of the illness. The participating physicians have referred to our surveillance system 590 patients who were ill during the first two years of surveillance with severe, debilitating, unexplained fatigue for at least the preceding six months. Referred patients were screened for psychiatric disorders preceding, concurrent with, and subsequent to the onset of their fatigue by specially trained nurses using a modified Diagnostic Interview Schedule. Complete health histories were obtained by interview and review of medical records and a basic panel of standard laboratory diagnostic tests were conducted. Four physicians have independently reviewed the health information of 337 of the patients for classification. Approximately 26% of patients referred to the surveillance system met the CFS case definition in all regards, 14% lacked one or more of the required eight symptom criteria, 15% were judged to have another possible or known medical illness which could account for the severe fatigue, and the remaining 45% did not meet the case definition because of histories of psychiatric disorders preceding the onset of fatigue. Minimum prevalence rates for the period 1 September 1989 to 1 September 1991 ranged from 2.0 to 7.3 per 100,000 of the general population across the four study sites and rates based on prorated data ranged from 4.6 to 11.3 per 100,000. More than 80% of the CFS cases were female, most were white, and their average age at onset was approximately 30 years.